Table 2 Micro-F1, macro-F1 and AUROC results for the prediction of COVID-19 in-hospital death.

	Macro-F1	Micro-F1	Precision-death	Recall-death	Log loss	AUROC
Stacking	0.654	0.821	0.562	0.354	6.032	0.826
LGBM	0.648	0.825	0.555	0.345	6.177	0.824
Lasso + population meta-features	0.633	0.816	0.550	0.319	6.355	0.794
STACKING + population meta-features	0.631	0.809	0.544	0.320	6.593	0.759
GAM	0.630	0.813	0.565	0.309	6.456	0.620
RF + population meta-features	0.626	0.816	0.581	0.299	6.338	0.811
LGBM + population meta-features	0.625	0.812	0.563	0.301	6.504	0.751
CNN1D	0.625	0.776	0.422	0.412	7.721	0.721
SVM + population meta-features	0.619	0.814	0.561	0.281	6.421	0.782
Resnet50	0.617	0.780	0.458	0.381	7.588	0.764
RF	0.617	0.817	0.584	0.275	6.317	0.809
GAM + population meta-features	0.616	0.817	0.580	0.279	6.323	0.609
FNet	0.611	0.779	0.439	0.350	7.642	0.720
SVM	0.608	0.814	0.574	0.255	6.424	0.813
LASSO	0.595	0.809	0.555	0.241	6.611	0.811
Resnet50 + RBF-kernel	0.593	0.752	0.383	0.371	8.577	0.698

List of model names from top to bottom (ordered by MacF1): CNN = convolutional neural network, FNet = fourier transformation neural network, FNet + VAT = fourier transformation neural network with virtual adversarial training, GAM = generalized additive models, KNN = K-nearest neighbors, LASSO = least absolute shrinkage and selection operator regression, LGBM = light gradient boosting machines, RF = random forest, SVM = support vector machines, Resnet50 = Residual Neural Network (with 50 residual blocks), STACKING = a stacking classifier, which combines all others.

Quick links

Search